AFAIK, the relinking requirement only applies if you distribute the software to someone.
If you’re running Sidequest entirely on your own infrastructure to orchestrate jobs across your backend, you’re not distributing the software at all, you’re providing a service. The tight coupling does not itself trigger extra obligations. What matters legally is distribution, not architecture.
Edgecase is if you give your software to a customer to run on their own servers (self‑hosted deployment/docker image shipped to customer). In those cases, you would need to allow them to replace Sidequest.js (ie, not obfuscating it away).
Someone more knowledgeable can correct me, if I'm wrong
Layman here also, but I think you are correct about GPL and LGPL (this case), but not for AGPL which adds a requirement that: "... If your software can interact with users remotely through a computer network, you should also make sure that it provides a way for users to get its source"
It appears from initial reading that it must be possible to support pure NLP tasks with this, but there weren't examples for these in the documentation, so I'm not sure. Does it support NLP models?
Ex: Could I have a store of articles and run NLP tasks against it?
Great question! Yes, EVA supports NLP pipelines thanks to our recent integration of Hugging Face pipelines last month. Here is an illustrative text classification application:
-- Text classification application in EVA
CREATE TABLE IF NOT EXISTS MyCSV (id INTEGER UNIQUE, comment TEXT(30));
LOAD CSV 'csv_file_path' INTO MyCSV;
CREATE UDF HFTextClassifier
TYPE HuggingFace
'task' 'text-classification';
SELECT HFTextClassifier(comment) FROM MyCSV;
EVA supports many other NLP pipelines, including summarization and text2text generation.
[2] is an illustrative notebook that presents an HF-based object segmentation pipeline (not NLP-based though). We would love to jointly explore how to best support your NLP pipeline. Please consider opening an issue with more details on your use case.
Do you have a pdf-only variant for the 12-pack? I looked around and didn't find it, I might have missed it. Hence asking. I don't really need printed version, so don't really want to waste paper.
I have been following your blogs and zines for years now. Keep up the good work and vibes. Peace.
A channel worth subscribing. Matt (and team?) manages successfully to walk the fine line between managing the complexity of the topics they discuss (which are sometimes extremely dense), making it consumable for the population that's interested in science and physics, but doesn't pursue it on a daily basis. At the same time, they don't fall for the trap of dumbing it down to the point where the audience develops misconceptions and starts believing in pseudo scientific claims.
There are some series they sometimes do, which are entertaining to follow. I also enjoy things like Journal club, where they pick a paper and deep dive into it. The audience also participates, in a way, where they pick the questions/comments from the previous video and answer them.
However, it's not all academic either. There are some running jokes etc, which keep the content entertaining, while being informative, a format that I see common in some of the best Youtube channels.
>making it consumable for the population that's interested in science and physics, but doesn't pursue it on a daily basis.
While I love the channel, it seems like they're more targeted to someone with a lot of physics education, more so than lay people. This is the only channel I've ever watched and genuinely felt dumb on a regular basis.
I think that's the key. Without a somewhat rigorous treatment, most interesting physics just doesn't work. The magic is that a well made resource is still consumable without quite understanding the complicated stuff. Then, if you come across something else later, maybe the commonalities make something click.
There's definitely a lot of material that, I have to accept that "I don't have the math for it"[0].
But Matt does a pretty good job of breaking things down conceptually, so that even though you don't know the specifics, the broader picture is at least vaguely understandable. Even if it means you need to watch previous episodes to get the gist of other discussed concepts.
[0] Harry Wilson, from John Scalzi's "Old Man's War"
Hah. I agree, but I'll add that his style is more like it's easy to follow for a while, then suddenly the car drives off a cliff and you need upper division college physics to understand.
I really love Anton. He has a way to ELI5 that's really effective and really genuine.
When we first started the show we were doing way dumber more youtube-y content (largely due to the interests of the producers, not PBS — look up the Majora's mask or farting in space episode). We got a ton of initial feedback early on that people wanted more actual science stuff, and it seemed like there was effectively no upper bound for what people would "tolerate" in terms of actual science vs. IFLS-core content. I think people honestly just like the challenge of understanding the show.
As someone with zero formal education in physics, I’ve always felt like they do a pretty good job. While there might be some parts that I have to dig into to understand or just accept going over my head, for the most part I come away from each video feeling like I understand both what they were talking about and what parts of what they were talking about were just surface level explanations.
I've watched random videos from it until I decided to watch them all chronologically. I must say, they do build on previous episodes, even more heavily as time goes on
Yeup, I took several 100 level physics in college, 400 level physical chemistry which involved quantum mechanics and got good grades on them. I can't understand the show.
Yes! If you watch more than 5 physics docs that run on TV, you quickly get tired of hearing "Black holes are so massive that not even light can escape it!".
I've loved Sublime for years, and it's always the first software I install on a new machine. I even voted with my pocket by buying a license.
However, since we started using Yarn workspaces (for JS), I've needed to switch to VSCode because its auto-import is just so much better, and it's one of those things that's hard to go back from, once you're used to it. Sublime text already indexes my code for search, which can probably be used for path suggestions / automatic imports without affecting performance too much maybe? I dunno.
I wouldn't mind it being a plugin either (before Yarn workspaces, I'd use FuzzyFilePath plugin which worked pretty well). But native support would mean the performance would be on par with what I've come to expect from Sublime Text :)
Wire is pretty popular in my social circles, less so lately, unfortunately.
The "always on" encryption is great, and it's always been more stable, and had a more complete feature set than Signal, even at launch. Most importantly, migration between devices with Wire is super smooth.
I don't know why you were downvoted. I'm the same; I still wear the clothes I got as swags when I was in college a decade ago. I've maybe bought a pair jeans since then, when my older ones didn't fit anymore.
I was able to enjoy TNG from the get go. It's one of the few shows where the stories have any depth. Although I did start watching it when I was in my twenties, so that could have been a factor in being able to pick up on the subtler themes
If you’re running Sidequest entirely on your own infrastructure to orchestrate jobs across your backend, you’re not distributing the software at all, you’re providing a service. The tight coupling does not itself trigger extra obligations. What matters legally is distribution, not architecture.
Edgecase is if you give your software to a customer to run on their own servers (self‑hosted deployment/docker image shipped to customer). In those cases, you would need to allow them to replace Sidequest.js (ie, not obfuscating it away).
Someone more knowledgeable can correct me, if I'm wrong